TensorAccessor

This podcast episode introduces TensorAccessor, a data structure in PyTorch that allows efficient element-by-element tensor manipulation, including for tensors with complex layouts, by providing direct access to strided data without contiguous memory allocation.

Outlines

Sign in to continue reading, translating and more.

Continue

PyTorch Developer Podcast

Tensor Accessor: Addressing Performance Costs in Tensor Computations

Efficient Data Manipulation in Tensors Using Raw Pointers

TensorAccessor: A Specialized Tensor for Strided Access

Tensor Accessor: A Powerful Tool for Efficient Tensor Manipulation

Tensor Accessor: A Powerful Tool for Efficient Tensor Manipulation

TensorAccessor

PyTorch Developer Podcast

00:00Tensor Accessor: Addressing Performance Costs in Tensor Computations

Tensor Accessor: Addressing Performance Costs in Tensor Computations

03:01Efficient Data Manipulation in Tensors Using Raw Pointers

Efficient Data Manipulation in Tensors Using Raw Pointers

03:48TensorAccessor: A Specialized Tensor for Strided Access

TensorAccessor: A Specialized Tensor for Strided Access

06:08Tensor Accessor: A Powerful Tool for Efficient Tensor Manipulation

Tensor Accessor: A Powerful Tool for Efficient Tensor Manipulation

09:32Tensor Accessor: A Powerful Tool for Efficient Tensor Manipulation

Tensor Accessor: A Powerful Tool for Efficient Tensor Manipulation